Testing does FX-83xx seem to flat line after 4.3GHz

[tungureanu]

I thought it was yo mamma

This?





Fixed!

 

[caddi daddi]

just a placeholder tungureanu

 

[tungureanu]

just a placeholder tungureanu

Placeholder may refer to:

In language:

Placeholder name, words that can refer to objects or people, whose names are unknown or irrelevant
Filler text, shares some characteristics of a real written text, but is random or otherwise generated
Lorem ipsum, text used to demonstrate the font, typography or layout in a document

In mathematics and computer science:.

Free variables and bound variables, symbols that will later be replaced by some literal string
Metasyntactic variable, alias term used to denote the subject matter under discussion
Format placeholder, used to format strings in print functions

In politics:

Placeholder (politics), a person appointed to an office temporarily, until a permanent replacement can be chosen.



pick one.

 

[Johan45]

This?

Not quite I was thinking yo mamma should've made you play outside more. Wither that or you're still in hyperspace.

 

[tungureanu]

both? more of the first part. but both

 

[Johan45]

Nice!!
I never had a choice. The old man needed help on the farm so....

 

[caddi daddi]

gawd, post #221 all i see is a place holder for the attachment.

 

[Johan45]

you don't see it in my response

 

[RGone]

I kant...

.

Kant see shett for the flashing lights.

Plus I always have to turn around and spread my legs and hands behind the head. How in hale U gunnuh see anythang?

RGone...ster.

 

[tungureanu]

.

Kant see shett for the flashing lights.

Plus I always have to turn around and spread my legs and hands behind the head. How in hale U gunnuh see anythang?

RGone...ster.

You should never stop because the popo says so. Just run like Bo and Luke. Like so.....

 

[caddi daddi]

it's fixed,

 

[RGone]

Nice pic there tungureanu.

 

[storm-chaser]

So after all the data collection, all the tests and various overclocking configurations plus benchmark results, what's your conclusion at this point? Does the FX 83XX seem to flat line after 4.3Ghz?

It appears from the information you've collected and based on my interpretation of the data, again I'm no expert, you are since this was your project from the start, but with the data in mind, it seems that the CPU does not flat line after 4.3Ghz. Do you agree, Rgone?

 

[caddi daddi]

I for one say it does, but not for the bench numbers.
it is the heat load that really kills the 8 core version after 4.3 or 4.4 when loading the cpu for hours on end.
I run swiftec rads on mine and they are rather on the thin side for it but even with the big rad the ability of the waterblock to tranfer the heat out of the cpu and into the water becomes the issue. for other than 100% cpu loads such as games that are video card intense, if you have the cooling it's a go, if you run thin rads and quiet fans it's pretty much a no go for me just for the fan noise.

 

[storm-chaser]

I for one say it does, but not for the bench numbers.
it is the heat load that really kills the 8 core version after 4.3 or 4.4 when loading the cpu for hours on end.
I run swiftec rads on mine and they are rather on the thin side for it but even with the big rad the ability of the waterblock to tranfer the heat out of the cpu and into the water becomes the issue. for other than 100% cpu loads such as games that are video card intense, if you have the cooling it's a go, if you run thin rads and quiet fans it's pretty much a no go for me just for the fan noise.

Earlier in the thread heat was mentioned as a limitation more than the CPU itself, but we now have Rgone running a high performance air cooler and reaching 4.6Ghz stable, so I agree with your reasoning that with the right cooling your FX will scale just fine.

 

[caddi daddi]

you are spot on storm-chaser, but I have found p95 to be a good test but not heavy enough for what i have used the rig for.
even ibt does not load it like my cfd computations do.

 

[Mandrake4565]

I really haven't found that the 8350's stop scaling it's more a matter of how much voltage you're willing to put to it and how much heat you can dissipate on a daily basis. Then again I do not do the kind of work on my pc like C/D does. Though I was able to run it for the entire Chimp Challenge 24/7 @ 4.7 without a problem and my temps were in the high 40's low 50's on the package/core 60's on the socket, with a 70 f ambient room. I am also using a 60mm wide radiator as opposed to the thinner Swiftechs C/D uses. I run 24/7 at 4.7 because 4.8 requires a significant jump in Cpu V to be stable from 1.46 to around 1.49 from what I remember. For benching the only thing stopping mine for now is heat, I've had it as high as 5475 @ 1.62 and I believe it will do more with better cooling.

 

[Johan45]

Earlier in the thread heat was mentioned as a limitation more than the CPU itself, but we now have Rgone running a high performance air cooler and reaching 4.6Ghz stable, so I agree with your reasoning that with the right cooling your FX will scale just fine.

I would just ike to add that RGone has a nice slice of silicon as well. I never ran mine on air but it needed 1.476v at 4.6 and I think that would be tough to handle even for a " good" air cooler.
These guys have pretty much summed the rest up.

 

[RGone]

My RE-think and thoughts today...

.
The idea of my testing to see if the FX-8 core PileDriver cpus acted as the FX-8 core BullDozer cpus, came from some testing that "ssjwizard" did with his FX-8120 vs his FX-8320. He also did a very good 'read' on his results.

The idea that has floated since the time of the initial FX release (among some) was that the struggle by most FX users to reach some great nirvana point of speed like 5.0Ghz that was the 'hyped' hot spot from most review sites might in fact be a myth of sorts.

Most of us that helped users from the initial BullDozer release, found the majority of users coming into the AMD Cpu Forum section to be woefully under-prepared to attain high FX CPU Mhz. Poor choice of motherboards and cooling that was just not up to the task of allowing super high Mhz for sure on FX 8 core processors to maintain a real semblance of stability. Run far less strenuous benches, perhaps, but an actual form of stress tested stability at high CPU Mhz, was doubtful. The 'hyping' review sites only ran benches and did not normally put the FX 8 core processors under something like a Prime 95 Blend test for any length of time.

Thinking back in time, it was the testing of "ssjwizard" comparing his performance between his FX-8120 and his FX-8320, that prompted my idea to test my PileDriver 8 core against my previous BullDozer 8 core processor. The prompting was to see if it was really such a good idea to begin the struggle to reach 4.6Ghz and even beyond to gain great gains in performance. A performance level not attainable by likely 85% of the users flooding into the AMD CPU Forum section with the request as always to help them overclock their FX processors on generally sub-par cooling and a motherboard not up to the task of outputting huge current to pushed 8 core processors.

At this point I chose the title of my earlier posted work as "Testing does FX-83xx seem to flat line after 4.3GHz" and that title might in retrospect been better if it were instead one like this >> Is It Really Productive to Push AMD FX 8 Core Processor Beyond 4.3Ghz CPU Speed. As I said in retospect, I might have chosen a different titling. But I did not and we are here today.

Now I will go slightly off topic here and present an idea that I now understand much better. Charts and Graphs can be used to support nearly any idea, depending on how the Chart or Graph is used to represent data inputs. No wonder so many try and support the less than great ideas with a Chart or a Graph. The Charts and Graphs can be misleading to a fault. People can look at a Chart or Graph and often draw a conclusion that is far away from what was intended to be the representation of the Chart or Graph. We know that many write into forums after buying parts and their question is always; "is my purchase worth a crap"? This after the deed is already done in buying. All they need is a Chart of Graph supporting their theory or purchase. Affirmation is huge in human beings. Data arrangement in a Chart or Graph can often be presented to represent something much different than what is actually in play in the actual overall scope of things. This happening was pure revelation to me. Back on topic.

The data I collected was done by incrementing the CPU speeds upward on a Bulldozer FX 8 core and then repeating the same speed incrementing upward on a PileDriver 8 core FX processor. The "data" itself was the result of running the CPU Test portion of Cinebench R11.5 in its' latest version at time of testing. Then I presented the Graphs to represent a percentage of "change" as each faster CPU Mhz was attained.

The initial Graph showed a hefty increase in scoring in the CPU performance test run on Cinebench R11.5 at 4.3Ghz. Good since many somewhat under-prepared users might attain that CPU Mhz even though not prepared for the HEAT of an 8 core FX processor when pushed with all 8 cores enabled as is the general situation when we overclock. Glancing at the first Graph representing the FX 8 core BullDozer processor, there was not anything like a linear gain in performance to come as a result of struggling in daily life to go beyond about 4.3Ghz. I repeat for emphasis that the FX Bulldozer Graph was intended to represent the lack of linear added preformance to be attained by pushing the CPU Mhz of the FX 8 core BullDozer faster and faster. Sure there was some actual test result increase, but not just upward and upward linearly in such a manner that I could expect a lot of "extra" performance for each 100Mhz increase in processor speed. Said another way the data seemed to indicate handily that just pushing and pushing a BullDozer FX 8 core had a point where it might well be counter-productive due to the need for so much CPU Voltage to stabilize the CPU as the speed increased; with the resultant HEAT to deal with. Short and sweet. Performance enough in daily use to make the effort to keep increasing processor speed; sort of a wasted exercise. Realizing this could allow many joe six pack users to rest more easily with their under-preparedness of components and the "hyped" idea that everyone of their FX processors was a dud since they were unable to reach the speeds 'hyped' in many reviews as FX nirvana. Help for aspects of human nature if you will.

Then the same sort of graphing for my FX-8350 put thru the same set of testing at 100Mhz increments in CPU Speed increase. The graph line from 4.2Ghz thru 5.1Ghz is a more regular looking sawtooth pattern. Looking at the direction of the graph line, I put in the text edit of the original picture that I had chosen 4.6Ghz as my normal daily speed for my own FX-8350 processor because after the 4.6Ghz mark the percentage increase in tested performance had certainly begun to drop off.

Another strange thing that I saw from the two graphs was that the high points in the graphing for the Bulldozer were on the odd numbered speed increases and the high points for the Vishera were appearing on the even numbered speed points. No clue as to why this might happen. In truth I really have no way of knowing why this might happen. Now in day to day practice I no longer even care as it has no real bearing on my speed choices for my CPU. What the PileDriver peaking on even numbered speeds allowed me to decide is that I could even run 4.2Ghz and have pretty darn good performance and power in my computing and this fact is a blessing in disquise after I slapped a HUmongoUS air cooler on my CPU for testing. I can fall back to 4.2Ghz for use and not be giving the shett house away and run cool and even quiet with the Air Cooler's fans turned down.

In rounding up my thoughts for closing this post out, I must admit that I have no idea how this whole thread has been used in peoples thinking about struggling so hard to reach high CPU speeds with the FX processors and now with today's about 12% better performing PileDriver, no matter its' actual processor speed. I have only a few reports from anyone outside this forum that have seemed to actually have viewed this testing exercise. In my mind after seeing the number of users still coming into the AMD CPU Forum section and beginning an overclock and wanting Max CPU Speeds, I think this testing is a bust. A test only for use by those that have inquiring minds and are past the extreme speed thing except for playing in general. I will say that the previos testing and results are beneficial to my own "human" mind when I swapped over to an Air Cooler for testing and still have my FX-8350 under air. I know for real that I have no need to beat up my CPU with extra voltage and HEAT to reach some high CPU speed for the type of use my own system is put to use doing.

Another point I came across in the testing I did is to know that you can take a Cinebench R11.5, CPU Test result that I did and half it for the performance of a 4 Core processsor and take 3/4s of it and it will represent the Cinebench R11.5 CPU Test for Performance of an FX 6 core processor. This is almost a given within 1 or 2 percent. So if I know my own CPU test performance outcome, I can determine what would be the outcome of another number of even cores at the same speed, within about a 2% margin. So when one of my buds asks me what I think of their Cinebench R11.5 scores, a little math and I am done with an answer that is related to my own results. Neat really.

Rounding up thoughts completely now; I have used my own testing to set more realistic expectations for my CPU speed struggle. I can rein-in my efforts to clock the ever-loving P out of my FX-8350 and not have to struggle constantly with AMD FX processor HEAT output.

I have come fully around to knowing that I can run much cooler, with less struggle and noise and do real live work quite handily at even as low as 4.2Ghz on a PileDriver cpu. This actually was what was first perceived when the initial testing was undertaken. I was actually trying to determine was the amount of performance increase as measured by Cinebench R11.5 worth the hassle to try and attain for a daily use setup? I settled in my own mind that I could realistically choose certain speeds that were producing added performance or not doing so and run my own processor at a speed that had a much more managable temperature.

Perhaps the thread title might have been; "is it worth the effort to go beyond 4.3Ghz with PileDriver as compared to the BullDozer cpu?" But the thread title has been what it is. This post is only done now to make perhaps better explanation of the results as I view them today.
RGone...ster.

Just for ease of reference without returning to the very first page, here is the Graph for the FX 8 core BullDozer I am referencing.

Here is the PileDriver Graph I was also referencing in text above.

 

[caddi daddi]

well said Mr. Rgone.